November 5, 2024 • 1hr 26min

Andrew Homan & Chris Miller - Redefining Semiconductor Progress - [Invest Like the Best, EP.395]

Invest Like the Best with Patrick O'Shaughnessy

This episode features Andrew Homan, Managing Partner at Maverick Silicon, and Chris Miller, professor at Tufts and author of "Chip War." They discuss the current state and future of the semiconductor industry, particularly in light of the AI revolution. The conversation covers everything from manufacturing dynamics to investment opportunities in the space.

0:00

-0:00

Key Takeaways

Paradigm Shift to AI: The semiconductor industry is undergoing a major architectural change with AI, potentially broader in implications than previous shifts like mobile and cloud computing
Infrastructure Layer Value: The bottom infrastructure layer (chips, cloud) is likely to capture significant economic value in the AI revolution, similar to how Intel dominated in PC era and ARM/Qualcomm in mobile
Power Dynamics: The hyperscalers (Meta, Google, Microsoft, etc.) are investing heavily in custom chip development to avoid becoming "dumb pipes" like telecom companies did during the mobile revolution
Manufacturing Concentration: TSMC's dominance in manufacturing comes from their singular focus and robust ecosystem in Taiwan, making it very difficult for other regions to catch up
Capital Requirements: While semiconductors require significant upfront capital, the margins can be more attractive than software at scale. Software requires more capital later in the lifecycle

Introduction

This episode features Andrew Homan, Managing Partner at Maverick Silicon, and Chris Miller, professor at Tufts and author of "Chip War." They discuss the current state and future of the semiconductor industry, particularly in light of the AI revolution. The conversation covers everything from manufacturing dynamics to investment opportunities in the space.

Topics Discussed

Intel's Historical Success and Current Challenges (00:06:28)

The discussion begins with Intel as a case study of how successful companies can struggle to adapt to industry changes. Intel's problems stem from their previous success in PC and data center chips, which made them risk-averse and hesitant to invest in new markets like mobile and AI.

Missed Opportunities: Intel famously declined to make chips for the iPhone
Institutional Challenge: Despite understanding the Innovator's Dilemma, companies struggle to sacrifice existing profitable business lines
Cultural Impact: Success can create internal resistance to promoting new business streams that threaten existing ones

The AI-Driven Paradigm Shift (00:08:22)

The industry is experiencing a major architectural shift toward AI, following previous transitions from mainframe to PC to mobile/cloud.

Scale of Change: The AI paradigm shift may be broader in implications than previous transitions
Business Model Evolution: Movement from integrated companies to fabless/foundry model, now seeing hyperscalers designing their own chips
Vertical Integration: Cloud companies are investing heavily in chip design to control the entire stack from software to silicon

Infrastructure Layer Economics (00:11:44)

The discussion explores where value will accrue in the AI stack, with particular focus on the infrastructure layer.

Three-Layer Model:
- Bottom: Chips and cloud infrastructure
- Middle: Foundation models (OpenAI, Anthropic, etc.)
- Top: Applications (ChatGPT, Copilot, etc.)
Value Capture: The infrastructure layer appears positioned to capture significant economic value regardless of which specific applications or models win

Manufacturing Dynamics and TSMC's Dominance (00:19:02)

TSMC's manufacturing leadership stems from multiple factors that make their position difficult to challenge.

Ecosystem Advantage: Taiwan's concentrated ecosystem of suppliers, expertise, and support infrastructure
Manufacturing Complexity: Chip production requires extreme precision and constantly evolving processes
Yield Challenges: The complexity of manufacturing means significant portions of chips produced may not work

Investment Landscape (00:29:28)

The semiconductor investment landscape has unique characteristics compared to software investing.

Capital Requirements: While upfront costs are high, margins can be better than software at scale
Founder Profile: Typically more experienced founders from large semiconductor companies rather than young entrepreneurs
Competition: Relatively few dedicated semiconductor investors, especially in mid-stage private markets

Edge AI and Future Applications (00:51:27)

The discussion explores opportunities beyond data center AI, particularly in edge computing.

Market Potential: Edge AI represents a largely untapped opportunity compared to data center AI
Power Efficiency: Edge applications require different optimization priorities, particularly around power consumption
Industrial Applications: Significant opportunity for AI deployment in industrial settings

Geopolitical Considerations (01:05:08)

The semiconductor industry is heavily influenced by geopolitical factors and government policy.

US Policy Goals:
- Reduce concentration of manufacturing in Taiwan
- Maintain US technological leadership
- Prevent adversaries from accessing cutting-edge technology
China's Challenges: Difficulties in developing domestic capabilities due to export controls and ecosystem limitations
Government Support: CHIPS Act providing $40B for US manufacturing, though this represents less than TSMC's annual capex

Future Outlook (01:10:57)

The conversation concludes with perspectives on the industry's future trajectory.

Market Size: Semiconductor industry projected to potentially exceed software market by decade's end
Innovation Areas: Memory architectures and chip-to-chip communication as potential breakthrough areas
Ecosystem Development: Need for continued development of startup ecosystem and talent pipeline

Conclusion

The semiconductor industry is undergoing a transformative period driven by AI, with significant opportunities for value creation in both infrastructure and edge applications. While the industry faces challenges around manufacturing concentration and geopolitical tensions, the fundamental drivers of growth remain strong. Success in this space will require deep technical expertise, significant capital, and careful navigation of political considerations.